Tgf-Β1 Transcriptionally Promotes 90K Expression

Grassadonia et al. Cell Death Discovery (2021) 7:86 https://doi.org/10.1038/s41420-021-00469-1 Cell Death Discovery

ARTICLE Open Access Tgf-β1 transcriptionally promotes 90K expression: possible implications for cancer progression Antonino Grassadonia1,VincenzoGraziano 1,2, Sara Pagotto 1, Angelo Veronese 3, Cesidio Giuliani3, Marco Marchisio3, Paola Lanuti3, Michele De Tursi1,MauriziaD’Egidio1, Pietro Di Marino1,DavideBrocco1,PatriziaVici4, Laura De Lellis5, Alessandro Cama5, Clara Natoli1 and Nicola Tinari1

Abstract The 90K protein, also known as Mac-2 BP or LGALS3BP, can activate the immune response in part by increasing major histocompatibility (MHC) class I levels. In studies on a non-immune cell model, the rat FRTL-5 cell line, we observed that transforming growth factor (TGF)-β1, like γ-interferon (IFN), increased 90K levels, despite its immunosuppressive functions and the ability to decrease MHC class I. To explain this paradoxical result, we investigated the mechanisms involved in the TGF-β1 regulation of 90K expression with the aim to demonstrate that TGF-β1 utilizes different molecular pathways to regulate the two genes. We found that TGF-β1 was able to increase the binding of Upstream Stimulatory Factors, USF1 and USF2, to an E-box element, CANNTG, at −1926 to −1921 bp, upstream of the interferon response element (IRE) in the 90K promoter. Thyrotropin (TSH) suppressed constitutive and γ-IFN-induced 90K expression by decreasing USF binding to the E-box. TGF-β1 was able to overcome TSH suppression at the transcriptional level by increasing USF binding to the E-box. We suggest that the ability of TGF-β1 to increase 90K did not result in an increase in MHC class I because of a separate suppressive action of TGF-β1 directly on the MHC class I gene. We propose that the increased levels of 90K may play a role, rather than in immune response, in the context of β

1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; the TGF- 1-induced changing of the cellular microenvironment that predisposes to cell motility and cancer progression. Consistently, analyzing the publicly available cancer patient data sets cBioPortal, we found that 90K expression directly correlated with TGF-β1 and USFs and that high levels of 90K were signiﬁcantly associated with increased mortality in patients affected by different types of cancer.

Introduction the expression of major histocompatibility complex The Transforming growth factor-beta (TGF-β) poly- (MHC) class I7. The importance of suppressing class I peptides regulate the growth, function, and immune levels is clear since TGF-β1-deficient transgenic mice properties of cells1. The prevalent role of TGF-β1 in the have increased MHC class I levels in many organs and immune system is to induce tolerance and immune sup- develop a rapid, wasting, autoimmune disease8. pression2, in part by inhibiting IL-2 production and NK The 90K protein, also known as Mac-2 BP or activity3,4, promoting the differentiation of naive T-cells LGALS3BP, is an oligomeric glycoprotein that has been into Treg via Foxp3 transcription factor5,6, and decreasing identified in the serum of patients with tumors or acquired immunodeficiency syndrome (AIDS)9. In contrast to TGF- β1, 90K activates the immune defense systems of the Correspondence: Antonino Grassadonia ([email protected]) organism10. In particular, human 90K enhances the in vitro 1Department of Medical, Oral and Biotechnological Sciences and Center for Advanced Studies and Technology (CAST), G. D’Annunzio University, Chieti, generation of cytotoxic effector cells (NK and LAK) from Italy peripheral blood mononuclear cells (PBMC)11; increases 2 Cancer Research UK Cambridge Institute, University of Cambridge, IL-2 production by PBMC11 and increases MHC class I Cambridge, UK 12 Full list of author information is available at the end of the article antigen expression in human breast cancer cells . Edited by Erwei Sun

© The Author(s) 2021, corrected publication 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a linktotheCreativeCommons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 2 of 13

® To better understand the physiological role of 90K, we U2OS cells (ATCC HTB-96TM), a human osteo- have characterized its function in rat FRTL-5 thyroid sarcoma cell line suitable for co-transfection experi- cells. These cells were chosen because they can maintain a ments,weregrowninDulbecco’sModified Eagle’s normal functioning behavior in culture: (i) near diploid, Medium with 4.5 g/l glucose, supplemented with 10% (ii) absence of tumor phenotype, (iii) growth and fetal bovine serum, 1% Pen/Strep and 1% L-glutamine immune-related protein expression under the control of (Merck, Darmstadt, Germany). TSH13. In FRTL-5 cells, we have previously reported that rat 90K, similarly to its human homologous, was increased RNA isolation, northern blot analysis, and RT-qPCR by γ-IFN and that TSH was able to decrease both con- For northern blot, RNA was isolated using the com- stitutive and γ-IFN-induced 90K synthesis14. We have mercial kit Quickprep mRNA Purification (Pharmacia also shown that rat 90K acted in an autocrine or para- Biotech, Uppsala, Sweden) and northern analysis was crine/exocrine manner to increase MHC class I and that it performed using Nytran nitrocellulose membranes was a component of an immune response aimed at con- (Merck, Darmstadt, Germany) as described14. Filters were verting non-immune cells to potential antigen-presenting hybridized with the rat 90K and GAPDH cDNAs; cells (APCs)14. This effect was triggered by the presence of radiolabeling of probes and hybridization (1 × 106 cpm/ foreign double-strand nucleic acids in the cytoplasm ml) were also as described14. Rat 90K cDNA was the (viral, bacterial, or self)14. full-length clone previously described14. As expected, in rat FRTL-5 thyroid cells TGF-β1 For reverse transcription-quantitative PCR (RT-qPCR), decreases expression of MHC class I15 and inhibits cell total RNA was isolated using QIAzol Lysis Reagent proliferation as well as the thyrotropin (TSH)-induced (Qiagen, Hilden, Germany) according to the manu- functions such as iodide uptake and thyroglobulin facturer’s instruction and reverse-transcribed with the production16,17. High Capacity cDNA Reverse Transcription Kit (Ther- Thus, given the antagonistic properties of TGF-β1 and moFisher Scientific, Waltham, Massachusetts, USA). RT- 90K within the immune system and their opposite effects qPCR was performed using the QuantiFast SYBR® Green on MHC class I regulation in FRTL-5 cells, we investi- PCR Kit (Qiagen, Hilden, Germany) using specific rat 90K gated the effect of TGF-β1 on the expression of 90K. primers, F: 5′-CTGAACAGTCTACAGAAAGCTTCG-3′ Surprisingly, we found that TGF-β1-increased 90K gene and R: 5′-GACCTGGAAGCCCAACCT-3′; and GAPDH expression. primers, F: 5′-TGGGAAGCTGGTCATCAAC-3′ and R: In the present study, we investigated the mechanism 5′-GCATCACCCCATTTGATGTT-3′. Melting curve by which this occurred in the attempt to explain the analyses were performed according to the specification of apparent paradox and better clarify the biologic function the CFX96 Touch™ Real-Time PCR Detection System of 90K. (Bio-Rad, Hercules, CA, USA). 90K mRNA expression was normalized to the endogenous reference GAPDH Material and methods using the 2-Δct method (User Bulletin #2, ThermoFisher Cells Scientific, Waltham, Massachusetts, USA). Rat FRTL-5 thyroid cells (Interthyr Corporation, Mar- ietta, OH, USA) were a fresh subclone (F1) with all In vivo labeling and immunoprecipitation properties described18. They were grown in 6H medium FRLT-5 cells in 5H medium were stimulated with 1 × − consisting of Coon’s modified F12 medium, 5% heat- 10 10 M TSH, 100 U/ml rat γ-IFN, or 5 ng/ml TGF-β1as treated, mycoplasma-free, calf serum, 1 mM nonessential noted above. Biosynthetic labeling of proteins was amino acids, and a six hormone mixture: bovine TSH (1 × performed during the last 12 h by maintaining the − 10 10M), insulin (10 μg/ml), cortisol (0.4 ng/ml), trans- cells overnight with 35S-methionine (50 μCi/ml) in ferrin (5 μg/ml), glycyl-L-histidyl-L-lysine acetate (10 ng/ methionine-free F12 Coon’s modified medium (0.5 ml/ ml), and somatostatin (10 ng/ml). Cells were fed every well) containing 1% dialyzed FCS (ThermoFisher Scien- 2–3 d, were passaged every 7–10 d, were diploid and were tific, Waltham, MA, USA). The spent medium was pre- between their fifth and 25th passage. Their doubling time cleared with rabbit preimmune serum and incubated with with TSH was 36 ± 6 h; without TSH, they did not pro- the polyclonal anti-90K antibody14, followed by protein liferate nor exhibit most thyroid-specific functions. As A-Sepharose (Merck, Darmstadt, Germany). The immu- noted, in some experiments cells were maintained in 5H noprecipitate was separated by SDS-PAGE and proteins − medium (no TSH) then stimulated with 1 × 10 10 M detected by autoradiography. TSH, 100 U/ml rat γ-IFN, or 5 ng/ml TGF-β1 for 24 h unless otherwise noted. Experiments carried out using Oligonucleotides and plasmids forskolin (5 μM) and IGF-I (100 ng/ml) instead of TSH USF expression plasmids (pUSF1 and pUSF2) were gen- and insulin, respectively, showed similar results. erated by cloning USF1 or USF2 DNA into pCMV-tag2A

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 3 of 13

vectors (Agilent Technologies, Santa Clara, CA, USA), as forward primer, and the sequence 5′-TAC CAA CAG previously described19. The rat 90K promoter-luciferase TAC CGG AAT GC-3′ as reverse primer. The fragment chimera, containing the fragment spanning position −1963 was subcloned in pGL3 basic vector (Promega, Madison, to −1772 bp of the 90K 5′-flanking region subcloned in a WI, USA) at the XhoI and HindIII insertion sites. pGL3 basic vector (Promega, Madison, WI, USA), was also described14. This region referred to as P2, corresponds to the Transient expression minimal promoter defined by sequence homology among Transient transfections in FRTL-5 cells used a DEAE rat, human, and mouse genes and is functionally relevant as procedure22 and were performed as described20. Trans- it contains the interferon and TSH response elements20. fections used 2 μg of the pGL3 basic luciferase reporter Using site-specific PCR mutagenesis procedure, we created a gene or equivalent molar amounts of pGL3 containing the P2 fragment with four base substitutions in the interferon P2 rat 90K constructs with or without IRE mutation response element (IRE) (−1912 to −1902 bp). This con- (P2IREM). pSVGH, 5 μg, was added to measure transfec- struct, indicated as P2IREM, has the IRE sequence changed tion efficiency. In some experiments, cotransfections were in tcA gCt GAA GCT and was also subcloned in a pGL3 performed using 3 μg of pUSF1, pUSF2, or the control basic vector (Promega, Madison, WI, USA). vector pCMV-tag2A. In other experiments, the same pro- Oligo I–II is the region spanning from −1963 to −1912 cedure was used to transfect cells with 2 μg of the SV40- bp in the rat 90K minimal promoter; Oligo II is from driven pGL3 luciferase reporter vector or equivalent molar −1963 to −1932: 5′-AGC CTT GTC TGC AGC CAA amounts of this vector containing Oligo I–II, Oligo II, or CCC CAG AGG CAG CC-3′; Oligo I is from −1937 to Oligo I with or without the above-indicated mutations. −1912 bp: 5′-GCA GCC TCC GTC ATG TGT TTT CTG After transfection, cells were maintained 3–6h in 6H GA-3′. CM3, CM2, CM1, TM1, TM2, and CEM are two- medium and then shifted to 5H medium with no TSH. base substitutions of Oligo I: CM3, 5′-GCA aCa TCC After 24 h in the 5H medium, cells were treated with TSH − GTC ATG TGT TTT CTG GA-3′; CM2, 5′-GCA GCa (1 × 10 10M), γ-IFN (100 U/ml), or TGF-β1(5ng/ml). TCa GTC ATG TGT TTT CTG GA-3′; CM1, 5′-GCA Luciferase activity was assayed after 24 h. Values were GCC TCa GgC ATG TGT TTT CTG GA-3′; TM1, 5′- normalized for total cell protein and growth hormone GCA GCC TCC GTC ATc TGT TTc CTG GA-3′; TM2, activity. The activity of the empty vector was considered as 5′-GCA GCC TCC GTC ATG TGT TTcaTG GA-3′; and the control value unless otherwise noted. CEM, 5′-GCA GCC TCC GTC ATG gaT TTT CTG GA- Transfections of pGL3_hu90K were performed in U2OS 3′. Lowercase and bold letters denote the substituted cells using Lipofectamine 2000 reagent (ThermoFisher nucleotides. The oligonucleotide containing the con- Scientific, Waltham, Massachusetts, USA). Briefly, 10,000 sensus E-box binding site (bold and underlined), indicated cells were co-transfected with the pGL3_hu90K or the as E, is 5′-GGA AGC AGA CCA CGT GGT CTG CTT-3; empty vector (150 ng), in the presence or not of pUSF1 or EM is a two-base mutation (lowercase) in the E-box of E, pUSF2 vector (150 ng) and the Renilla luciferase reporter 5′-GGA AGC AGA CCA CGg aGT CTG CTT-3′. The vector pRLTK (15 ng) (Promega, Madison, WI, USA). cAMP response element (CRE) competitor was 5′-AGA After 24 h, firefly luciferase activity was measured and GAT CTG ACG TCA GAC AGC TAG-3′; the FAST-1 normalized to that of Renilla using the Dual-Luciferase binding element (FBE) competitor was 5′-CTG CCC TAA Reporter Assay System according to the manufacturer’s AAT GTG TAT TCC TAG GAA ATG 3′. The double- instructions (Promega, Madison, WI, USA). strand oligonucleotides used for gel mobility shift assay were obtained by annealing each oligonucleotide with its Nuclear extracts complementary antisense strand. To prepare nuclear extracts, FRTL-5 cells were grown Oligo I–II, Oligo II, or Oligo I, with and without in a complete 6H medium until 50–70% confluent. Cells mutations, were also subcloned into a pGL3 promoter- were shifted in 5H medium with no TSH for 5 days and − luciferase reporter gene vector containing the then stimulated with γ-IFN (100 U/ml), TSH (1 × 10 10 SV40 minimal promoter (Promega, Madison, WI, USA). M), TGF-β1 (5 ng/ml), or combinations thereof for the To construct the human 90K minimal promoter- times noted, usually 24 h. Nuclear extracts were prepared luciferase chimera (pGL3_hu90K) the fragment span- as described20. Protein concentration was determined by ning from −115 to +42 of the human 90K 5′-flanking Bradford’s method (Bio-rad, Hercules, CA, USA) with region (sequence number based on defining transcrip- crystalline bovine serum albumin (BSA) as standard. The tional start site as +1) was generated by PCR using the full extracts were aliquoted and stored at −70 °C. length 90K promoter present in the pBluescript II KS (+) vector (a gift of Prof. Axel Ullrich, Max Plank Institutes, Electrophoretic mobility shift assay (EMSA) Munich, Germany)21 as template, the sequence 5′-CCG DNA probes were the oligonucleotides described above. CTC GAG TGG GGA GTA TCA GCA GCA G-3′ as They were 5′ labeled with [γ-32P]ATP using T4

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 4 of 13

polynucleotide kinase (New England Biolabs). Binding Materials reactions, 20 μl in volume, included 5 μg nuclear extract, Highly purified bovine TSH was obtained from the 2% glycerol, 50 ng/ml polydI-dC, and 2 mM DTT in hormone distribution program of the National Institute of Binding Buffer: 20 mM HEPES, 0.1% NP40, 5 mM MgCl2, Diabetes and Digestive and Kidney Diseases, National and 50 mM KCl. Where indicated, specific antisera Institutes of Health (NIDDK-bTSH I-1; 30 U/mg). Rat γ- (0.4 ng) or unlabeled oligonucleotide competitors (20× or IFN was from Amgen (Thousand Oaks, CA, USA); 50× the concentration of labeled probe) were added for recombinant IGF-I was from the Fujisawa Pharmaceutical 15 min at room temperature. Reactions were initiated by Co. (Chūō, Tokyo, Japan). Antibodies against USF1 or adding 50,000 cpm labeled DNA probe; after 15 min at USF2 were from Santa Cruz Biotechnology, Inc. (Dallas, room temperature, samples were electrophoresed at TX, USA). Human platelet TGF-β1 was from Merck 160 V on 5% native polyacrylamide gels in 1×TBE at room (Darmstadt, Germany). [γ-32P]ATP (3000 Ci/mmol) was temperature. Gels were dried and autoradiography from Amersham (Arlington Heights, IL, USA). The performed. source of other materials was Merck (Darmstadt, Ger- many) unless otherwise noted. USF1 production in Escherichia coli Recombinant USF1 was produced using the pET system Results (Merck, Darmstadt, Germany). The USF1 cDNA was TGF-β1-increased 90K RNA levels and protein secretion subcloned into the EcoRI site of the expression vector Although FRTL-5 cells maintained in medium without pET-30(+), allowing the His-Tag sequence to be linked to TSH (5H medium) had a considerable expression of 90K its N terminus. The protein was produced in E. coli BL21 mRNA detectable by RT-qPCR (Fig. 1A, bar 1), the (DE3) after a 4-h stimulation with 1 mM isopropyl-β-D- addition of 5 ng/ml of TGF-β1 to the medium caused a thiogalactopyranoside (IPTG) and purified according to significant increase in 90K mRNA levels (Fig. 1A, bar 2 vs − the manufacturer’s procedures under reducing conditions. 1). When 1 × 10 10M TSH was added to the 5H medium, Purified USF1 was dialyzed against 20 mM HEPES, 0.1% there was, in contrast, a significant decrease in 90K NP40, 5 mM MgCl2, 50 mM KCl, 2% glycerol, 2 mM mRNA levels (Fig. 1A, bar 4 vs 1), as previously repor- DTT, and concentrated in a Centricon 10 (Amicon, ted14. The addition of TGF-β1 overcomes the suppressive Beverly, MA, USA) to be used in EMSA. The purity action of TSH, increasing 90K RNA levels to the same and integrity of the recombinant protein were verified by extent of that observed when TGF-β1 was added to 5H SDS/PAGE. medium (Fig. 1A, bar 5 vs 2). The increase in 90K mRNA induced by TGF-β1 was as Analysis of public platforms for ChIP-seq and clinical data prominent as that induced by γ-IFN (Fig. 1A, bar 3 vs 1 To confirm the binding of USFs in the promoter and bar 6 vs 4), particularly in 6H cells because TSH, in region of the human 90K gene, we analyzed the publicly the presence of insulin, reduces the γ-IFN effect, as pre- available ChIP-seq (chromatin immunoprecipitation viously reported14. RT-qPCR data were confirmed by followed by high-throughput sequencing) database northern blot analysis (see Fig. 1SA, B in Supplementary generated by the ENCODE consortium, freely accessible Materials). at https://www.encodeproject.org/.TheChiIP-seq Autoradiography of [35S]methionine-labeled proteins experiments were carried out on the human embryo- secreted by FRTL-5 cells and immunoprecipitated with a nic cell H1. specific anti-90K antibody revealed the presence of a high The cBioPortal platform, publicly available at https:// molecular weight form of 90K (~200 kDa) and one of www.cbioportal.org/, was analyzed to correlate 90K (both ~57 kDa, when cells were grown in 5H medium (Fig. 1C, mRNA and protein) with TGF-β1 and USFs expression lane 1). High molecular weight dimers of 90K protein and to investigate the relationship between 90K levels and have been identified by western blotting in studies on overall survival (OS) in patients affected by different types human fluids11 and the mechanisms involved in dimer- of cancer. ization or oligomerization have been detailed23. The 200- and 57-kDa bands are likely corresponding to 90K dimers Statistical significance and protein degradation products, respectively. In some All experiments were repeated at least three times with experiments, in 5H medium, a tiny band at ~90 kDa was different batches of cells. Values were the mean ± SD of also appreciable, corresponding to the 90K monomer14. these experiments where noted. Significance between Cells incubated with γ-IFN or TGF-β1 for 24 h (Fig. 1C, experimental values was determined by a two-tailed Stu- lane 2 and 3, respectively) secreted a prominent ~90 kDa dent’s t-test or two-way analysis of variance and was form, as well as an increased amount of the high significant if p values were <0.05 when data from all molecular weight 200 kDa form of the 90K protein. experiments were considered. The ~57 kDa form appeared increased, albeit slightly

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 5 of 13

Fig. 1 Effect of TGF-β1 on rat 90K mRNA and protein secretion in FRTL-5 cells. In A, RT-qPCR analyses to characterize the effect of TGF-β1 and γ-IFN on 90K mRNA expression in the presence or absence of TSH. In B, autoradiography of biolabeled secreted proteins immunoprecipitated by specific anti-90K serum. A single asterisk (*) denotes a significant decrease (p < 0.05); two asterisks (**) or three asterisks (***) denote a significant increase (p < 0.01 and p < 0.001, respectively). Data represent the mean ± SD of duplicate values determined in three separate experiments performed on different batches of cells. decreased in size. Moreover, an additional 47 kDa form Fig. 1B). Based on the activity of the P2 construct, we appeared in the medium from γ-IFN-treated cells (Fig. 1C, concluded that TGF-β1 acted on the same minimal pro- lane 2). The increase in 90K secretion (monomer and moter that was sensitive to γ-IFN and TSH. This region dimer) was consistent with the effect of TGF-β1, as well as includes the IRE and corresponds to the minimal pro- γ-IFN, to increase mRNA levels and protein synthesis, moter defined by sequence homology between rat, while that of 57- and 47-kDa forms was consistent human, and mouse genes20. with increased degradation of 90K in association with The insertion of 4 bp mutations in the IRE of the P2- increased synthesis and secretion. luciferase construct (P2IREM) (Fig. 2, upper diagram) The addition of TSH to cells in a medium containing abolished the ability of γ-IFN to increase promoter insulin (6H vs 5H), significantly decreased secretion of the activity (Fig. 2B, bar 2 and 6), but did not influence sig- ~200 kDa form of the protein and increased the amounts nificantly the effect of TGF-β1 or TSH (Fig. 2B, bar 7 and of lower molecular weight forms (Fig. 1C, lanes 4 to 6). 5, respectively), indicating that TGF-β1 activity was not This was consistent with increased degradation of 90K in mediated by the IRE. the presence of TSH, whether synthesis was increased by The sequence of P2 upstream of the IRE, termed Oligo TGF-β1orγ-IFN. Thus, TSH does not affect TGF-β1- I–II, was inserted into a pGL3 vector with a SV40 pro- induced 90K RNA levels but increases 90K protein moter (Fig. 2, lower diagram). In 5H medium, the Oligo degradation. I–II construct had a significant constitutive activity compared with its control vector without the insert Identification of the TGF-β1 responsive element in the (Fig. 2C, bar 1). TSH significantly decreased constitutive minimal 90K functional promoter activity (Fig. 2C, bar 2). TGF-β1, but not γ-IFN, overcame We measured the activity of the rat 90K minimal pro- the TSH-induced decrease in promoter activity (Fig. 2C, moter by luciferase assay as described previously20, using bar 4 vs 3). The Oligo I–II region was, therefore, the TSH the construct containing the fragment spanning position and TGF-β1 site of action. −1963 to −1772 bp of the 90K 5′-flanking region (P2 By inserting Oligo I wild type (−1937 to −1912) or Oligo construct) (Fig. 2, middle diagram). After transfection, I with 2 bp mutations into a pGL3 promoter, we had pre- FRTL-5 cells were maintained in 5H medium and sti- viously determined the site for constitutive 90K expression mulated with 5 ng/ml TGF-β1, 100 U/ml γ-IFN, or 1 × and TSH suppression20. Similarly, we could localize the − 10 10 M TSH (Fig. 2A). We observed that the luciferase TGF-β1 action within this region. Coordinate results activity of the construct in response to TGF-β1orγ-IFN were obtained when vector constitutive activity was eval- mimicked the effects of both on RNA levels (Fig. 2Avs uated simultaneously for TGF-β1- or TSH-responsiveness.

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 6 of 13

Fig. 2 Ability of TGF-β1 to increase promoter activity of rat 90K in FRTL-5 cells. In the upper diagram, the sequence of the rat 90K minimal promoter, P2 fragment, or the same construct with 4 bp mutations in the interferon response element (IRE), P2IREM, are depicted as inserted in a pGL3 basic luciferase reporter vector. The IRE is boxed to show the mutated nucleotides (lowercase letters). Luciferase activity was normalized for protein and growth hormone activity. In A, luciferase activity of the P2 construct. In B, luciferase activity of the P2IREM construct. In the lower diagram, the region between −1963 and −1912 bp upstream of the IRE, termed Oligo I–II, or the region between −1937 and −1912, termed Oligo I, or 5 different constructs with 2 bp mutations, were subcloned in a pGL3 promoter-luciferase reporter vector. Luciferase activity was expressed as relative to a control construct with no insert. In C, luciferase activity of Oligo I–II. In D, promoter activity of CM1 and TM1 is lost compared to Oligo I. In E, promoter activity of CM3, CM2 and TM2 is similar to that of Oligo I. A single asterisk (*) denotes a significant decrease in promoter activity (p < 0.05); two asterisks (**) or three asterisks (***) denote a significant increase in promoter activity (p < 0.05 and p < 0.01, respectively), a dark circle (●) denotes a significant decrease in constitutive promoter activity (p < 0.05).

In particular, as was the case for TSH-suppressive activity, activity involved the sequence spanning −1929 to −1917 the increased response to TGF-β1 was present in Oligo I bp, and did not involve the 5′ nucleotide mutation in construct, but was lost, along with constitutive activity, in CM1 (C-1929A) and the 3′ nucleotide mutation in TM1 the constructs containing mutations between −1929 and (T-1917C). −1917 bp (CM1 and TM1) (Fig. 2D). The promoter activity We preliminarily interpreted these data as follows. An exhibited by Oligo I and CM2, CM3, TM2 mutants were element important for the constitutive activity of the rat similar, as was their TGF-β1 and TSH response (Fig. 2E). 90K minimal promoter existed between −1929 and Since mutant CM1 lost both activities as did TM1, but not −1917 bp. This same element was also important for both TM2, CM2, nor CM3, it was evident that the critical resi- TGF-β1-increased and TSH-decreased 90K activity. dues for constitutive activity and for both TGF-β1orTSH EMSA data below support this interpretation and deﬁne

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 7 of 13

the cis element as an E-box and the trans factors involved important for constitutive as well as TGF-β1-increased as USF1 and USF2. 90K activity of the rat 90K minimal promoter was an E- box with the sequence CATGTG, consistent with the An E-box and USF transcription factors regulate known E-box site CANNTG. constitutive and TGF-β1-induced expression of rat 90K The E-Box is known to bind the basic helix-loop-helix As previously reported20, we performed EMSA to eval- zipper proteins, upstream stimulatory factors 1 and 2 uate the effects of the CM1, CM2, CM3, TM1, and TM2 (USF1 and USF2)24. When antisera to USF1 and USF2, mutations on trans factor binding to the Oligo I region but not a control serum, were included in the assay, the defined functionally as important for constitutive as well native complex was eliminated (low arrow) and super- as TGF-β1-enhanced or TSH-suppressed 90K activity shifted complexes appeared (Fig. 3D, lanes 9 and 10 vs 8). (Fig. 3A). A prominent protein/DNA complex was noted The increase in density of the complex, independent of incubating radiolabeled Oligo I with nuclear extracts of the sera added, was presumed to be a nonspecific salt cells maintained in 5H medium (Fig. 3A, lane 1). This effect of the antisera buffer (Fig. 3D, lanes 8 to 10). Anti- complex was nearly eliminated in incubations with radi- USF2 generated two supershifted complexes, the upper of olabeled CM1 or TM1 oligonucleotides (Fig. 3A, lanes 2 which was most prominent (Fig. 3D, lane 10, arrows). and 5, respectively) but was unaltered using radiolabeled Anti-USF1 appeared to generate a major supershifted CM3 or CM2 oligonucleotides (Fig. 3A, lanes 3 and 4). The complex with the same mobility as the faster migrating complex was only partly attenuated in incubations with one supershifted by anti-USF2, and a smaller amount of radiolabeled TM2 oligonucleotide (Fig. 3A, lane 6). the upper complex (Fig. 3D, lane 9, arrows). More importantly, formation of the complex was The ability of anti-USF1 and anti-USF2 to generate a increased in extracts from cells treated with 5 ng/ml TGF- supershifted complex with the same mobility and to β1 (Fig. 3B, lane 3 vs 2), but decreased in extracts from completely eliminate the native complex suggested that cells treated with TSH (Fig. 3B, lane 4 vs 2). the latter was represented by a USF1/USF2 heterodimer. The sum of these data suggested that the presence of The differences in the pattern of supershifted complexes this complex was important for constitutive 90K gene obtained with anti-USF1 and anti-USF2 suggested, how- expression, was increased by TGF-β1, but was decreased ever, that USF homodimers might also be interacting with by TSH. the E-box of the 90K gene in extracts from TGF-β1- In reviewing the sequence of Oligo I, we noted a treated cells. potential cyclic AMP responsive element (CRE) but found Using Oligo I or the unrelated EWT oligonucleotide no inhibition of the complex by an oligonucleotide with a (Fig. 3E, EWT) as radiolabeled probes, we found com- consensus CRE (Fig. 3C, lane 3). Similarly, we noted a plexes with identical mobility in EMSA with nuclear potential FAST-1 binding site but found no inhibition of extracts from TGF-β1-treated cells (Fig. 3E, lanes 2 and the complex by an oligonucleotide with a consensus 5). The same-sized complexes were also formed with FAST-1 binding element (FBE) (Fig. 3C, lane 4). The recombinant USF1 (Fig. 3E, lanes 4 and 6). More impor- exclusion of FAST-1 as a potential binding site was tantly, the addition of full-length pUSF1 or pUSF2 to important given the extent of homology and given that transfections enhanced the constitutive expression of P2, FAST-1 is involved in TGF-β1 signaling1. eliminated the TGF-β1-induced increase, and nearly We noted, however, a potential E-Box CANNTG at completely attenuated the TSH-induced suppression of −1926 to −1921 bp, and could show that binding of the constitutive P2 expression (Fig. 3F). This was not dupli- constitutive complex was not duplicated by an Oligo I cated by the control plasmid. with a critical mutation in the potential E-box (CEM), Given the high sequence homology between the rat and CATGga (mutation in bold and small letters), as was the human 90K minimal promoters, including the same loca- case for the CM1 oligonucleotide (Fig. 3D, lanes 3 and 2, tedIREandE-box20, we would verify if USFs were able to respectively, vs lane 1). Formation of a complex with the induce the human 90K gene. Co-transfection of U2OS cells same mobility was obtained with a radiolabeled oligonu- with pUSF1 or pUSF2 expression plasmids in the presence cleotide with a consensus E-box (CACGTG) and no other of pGL3_hu90K significantly enhanced the luciferase sequence homology to Oligo I (EWT) (see Fig. 3E, lines 5 activity of the reporter vector (Fig. 3G), indicating a direct and 6), but did not with the same oligonucleotide carrying effect of USF in inducing human 90K gene expression. the E-box mutation CACGga (EM) (Fig. 3D, lane 4). When used as competitors, both oligonucleotides, Oligo I USFs bind the E-box in the promoter region of the human and EWT, were inhibitory at the same concentrations 90K gene (Fig. 3D, lanes 5 and 7 vs 1). In contrast, there was no The ability of USFs to bind the E-box present in the inhibition when EM was used as competitor (Fig. 3D, lane human 90K gene promoter was confirmed by analyzing 6). These data supported the conclusion that the element the ChIP-seq database generated by the ENCODE

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 8 of 13

Fig. 3 (See legend on next page.) consortium. By selecting the human embryonic cell H1 as target gene, a peak region in the LGALS3BP promoter Biosample and browsing the platform for USF1, as a sequence of Chromosome 17 was identiﬁed in six differ- transcription factor, and human 90K (LGALS3BP), as the ent experiments (Fig. 4, lower diagram). As expected, the

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 9 of 13

(see figure on previous page) Fig. 3 Identification of a specific complex binding to the rat 90K minimal promoter important for TGF-β1-induced or TSH-suppressed expression of the 90K gene. In A, EMSA of nuclear extracts from cells maintained in 5H medium (no TSH) using as probes the radiolabeled double- strand oligonucleotides with the sequence of Oligo I or the ones with 2 bp substitutions reported in the lower diagram of Fig. 2.InB, EMSA of nuclear extracts from cells treated with TGF-β1 or TSH and incubated with Oligo I as a probe. In C, EMSA of nuclear extracts from cells treated with TGF-β1 using Oligo I as probe and each of the double-strand oligonucleotides depicted in the bottom as competitor of complex formation, i.e., Oligo I, with the potential CRE or FAST-1 (FBE) binding sites boxed, Oligo CRE, an unrelated oligonucleotide containing the consensus sequence for CRE, and Oligo FBE, an unrelated oligonucleotide with the consensus sequence for FBE. Nucleotides in Oligo I that are the same as the CRE are noted with dark circles; nucleotides in Oligo I that are the same as the FBE are noted with asterisks, In D, EMSA of nuclear extracts from TGF-β1-treated cells incubated with the indicated probes, competitors (20× concentration over labeled probe) or antisera (control, anti-USF1 or anti-USF2). CEM is OligoI with 2 bp mutations in the E-box; EWT is an oligonucleotide with a consensus E-box but no other sequence homology to Oligo I; EM is an unrelated oligonucleotide with a mutation in the E-box. The lowest arrow on the right denotes the native complex; the two upper arrows denote the supershifted complexes. In E, EMSA of nuclear extracts from TGF-β1-treated cells incubated with Oligo I or EWT used as probes. In lanes 4 and 6, they were respectively incubated with recombinant USF1 (rUSF1). In F, rat FRTL-5 cells were co-transfected with the pGL3 vector containing the P2 construct of the rat 90K gene and the plasmid containing the full-length cDNAs encoding USF1 (pUSF1) or USF2 (pUSF2) or the control vector pCMV- tag2A (control plasmid) or nothing (no plasmid). pSVGH was added to measure transfection efficiency. Values were normalized for cell protein and growth hormone activity. In G, human U2OS cells were co-transfected with pGL3 control vector or the pGL3 vector containing the human 90K minimal promoter (pGL3_hu90K), plus the Renilla luciferase reporter vector pRLTK to measure transfection efficiency, and with pUSF1 or pUSF2 or control vector as in F. Values were normalized for cell protein and Renilla activity, and expressed as relative to pGL3 control vector with no insert. A single asterisk (*) denotes a significant TSH-induced decrease in constitutive promoter activity (p < 0.05); two asterisks (**) denote a significant increase in promoter activity induced by TGF-β1 or by transfection with pUSF1 or pUSF2 (p < 0.01).

highest point of the peaks overlapped with the consensus curves showed that the group HH had significantly motif CANNTG binding site for USF (Fig. 4, upper shorter survival compared with the group LL (median diagram). disease-specific OS 40.44 months vs not reached, p < 0.001) (Fig. 5D). Importantly, in tumor samples, the Human 90K (LGALS3BP) levels directly correlate with expression levels of LGALS3BP showed a direct correla- TGF-β1 and USFs expression and affect cancer patient tion with those of TGF-β1(R = 0.34, p < 0.001), USF1 prognosis (R = 0.30, p < 0.001), and USF2 (R = 0.37, p < 0.001) The analysis of different cancer data sets included in the (Fig. 5E). In addition, in cancer cells, there was a strong public cBioPortal platform revealed a general association tendency of co-occurrence between high LGALS3BP, high between high tumor levels of LGALS3BP (both mRNA USF1, high USF2, and high TGF-β1 expression (p < 0.001) and protein) and shorter survival in patients affected by or low LGALS3BP, low USF1, low USF2, and low TGF-β1 different types of cancer. The effect reached statistical expression (p < 0.001) (Fig. 5F). Overall, clinical data significance in some tumors such as breast cancer, col- support a direct relationship between the activation of the orectal cancer, and clear cell renal cell carcinoma TGF-β1/USFs/90K axis and poor prognosis. (ccRCC). The latter neoplasm showed the highest statistical power in distinguishing prognosis between patients Discussion with high and low LGALS3BP levels. In particular, in a In this study, we showed that TGF-β1 increases 90K cohort of 512 patients with ccRCC included in the TCGA gene expression by increasing USF1 and USF2 binding to PanCancer Atlas database, the median disease-specificOS its promoter, in the same E-box involved in the hormonal was significantly shorter in patients with high LGALS3BP regulation of the gene. (EXP > 1) compared to those with low LGALS3BP (EXP < In fact, we previously demonstrated that USF1 and 1) (56.68 months vs not reached, p < 0.001) (Fig. 5A). USF2 were the relevant transcription factors controlling Similar results were observed among patients with high constitutive 90K expression and that TSH was able to (EXP > 1) and low (EXP < 1) TGF-β1 (median disease- decrease 90K expression by reducing USFs binding to the specific OS 53.42 vs 116.84 months, p < 0.001) (Fig. 5B), E-box20. and among patients with high (EXP > 1) or low (EXP < 1) Taken together, this and our previous report suggest USF1 (median disease-specific OS 56.68 vs not reached, that 90K gene expression depends on the interaction of p < 0.001) (Fig. 5C). Importantly, the magnitude of the the E-box located in its promoter with USF1 and USF2, effect on survival was significantly increased when and that this interaction is regulated by TGF-β1and patients with concomitant high LGALS3BP and high TSH: TGF-β1 increments USF binding to the E-box TGF-β1 (HH) were compared with those with low and increases 90K gene expression, while TSH reduces LGALS3BP and low TGF-β1 (LL), indicating a synergistic USF binding to the E-box and decreases 90K gene effect between the two genes (Fig. 5D). Kaplan–Meier expression.

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 10 of 13

Fig. 4 Genomic ChIP-seq reveals that USFs bind the E-box present in the promoter region of the human 90K (LGALS3BP) gene. ChIP-seq analysis was performed in the publicly available ENCODE platform. Search parameters were the human embryonic cells H1, USF1 as a transcription factor, and LGALS3BP as the target gene. Six different experiments showed a peak region in the LGAL3BP promoter in Chromosome 17, automatically aligned. The platform provided also the number of nucleotide position in the chromosome. The highest point of the peaks overlapped with the consensus motif CANNTG binding site for USF, depicted on the top. The arrow indicates the direction of gene transcription.

A role of USFs in the TGF-β1 signaling has been already remains to be defined, even though Smad-independent suggested in studies on the plasminogen activator pathways have been described35,36. – inhibitor-1 (PAI-1) gene25 28, also known as SERPINE1, The effect of TGF-β1 to increase 90K gene expression is but so far no other TGF-β-induced gene has been quite surprisingly given that TGF-β1 is a suppressor of the reported to be dependent on USF activity. The present immune system2,3 and decreases MHC class I expression report unequivocally, therefore, establishes the impor- in FRTL-515, while 90K is an immune-stimulator and tance of USF family transcription factors in the TGF- increases MHC class I. The finding that USFs are directly β1 signaling. involved in the TGF-β1 regulation of 90K gene provides a Classically, TGF-β1 signaling is mediated by receptor- mechanistic explanation: TGF-β1 suppresses MHC class I associated Smad2 and Smad3 proteins which form com- while increasing 90K because different transcription fac- plexes with Smad4 and accumulate in the nucleus where tors are involved in the regulation of the two genes. In regulates gene expression29,30. Smad2, Smad3, or Smad4 fact, in FRTL-5 cells, TGF-β1 has been shown to decrease can mediate TGF-β1-induced transcription by binding MHC class I gene expression by regulating diverse trans directly to DNA in specific Smad Binding Element factor interactions with two elements located within the (SBE)31,32. No SBE site exists in 90K minimal promoter, “hormonal regulatory region” of its promoter (−203 to but we cannot exclude an interaction of USFs with Smads −90 bp)15. The first element, called Enhancer A (−180 to to act cooperatively elsewhere in the intact 90K gene. −170 bp), can interact with either an inducer factor, the Smads can cooperate with other transcription factors and/ heterodimer NF-kB p50/fra-2 (MOD-1), or an inhibitor or can interact with co-activator such as FAST-1 to factor, the heterodimer NF-kB p50/p65. TGF-β1 reduces activate transcription33,34. Since a potential FAST-1 con- the interaction of MOD-1 with Enhancer A while sensus sequence was present in 90K minimal promoter, increasing that of NF-kB p50/p65. Both, reduced MOD-1 we specifically excluded FAST-1 as a mediator of the and increased p50/p65, suppress class I expression15. The TGF-β1 action on 90K gene. The role of the Smad system second element, called downstream regulatory element in the TGF-β1 regulation of 90K gene described herein (DRE) (−127 to −90 bp) is the binding site of the

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 11 of 13

Fig. 5 Human 90K (LGALS3BP) expression is associated with poor prognosis and directly correlates with TGF-β1 and USFs. Using the public cancer database cBioPortal, the TCGA PanCancer Atlas cohort of patients affected by clear cell renal cell carcinoma (ccRCC) was analyzed (N = 512). In A, Kaplan–Meier curves of disease-specific OS stratified by high (EXP > 1) or low (EXP < 1) expression of LGALS3BP. In B, patients were stratified by TGF-β1 expression; in C, by USF1 expression; and in D, by concomitant LGALS3BP/TGF-β1 high or low expression. In E, scatterplot correlation between LGALS3BP and TGF-β1 (left), LGALS3BP and USF1 (middle); LGALS3BP and USF2 (right), as visualized by cBioPortal. In F, table elaborated by cBioPortal showing tendency of co-occurrence among 90K, TGF-β1, and USF1. inhibitor factor TSEP-1, a ubiquitously-expressed Y-box Similarly, TGF-β1 is increased in most human cancer, protein. TGF-β1 increases the binding of TSEP-1 to DRE and high expression levels correlate with more and, as a consequence, suppresses class I expression15. advanced stages of malignancy and decreased survi- – These interactions appear to counteract or dominate the val42 44. The pro-metastatic effect of TGF-β1ismostly inductive effect of 90K to increase class I. related to its ability to remodel tumor microenviron- – In addition, a functional explanation can be provided ment45 48 and to promote epithelial–mesenchymal for the effect of TGF-β1 on MHC class I and 90K: the transition (EMT)49,50. upregulation of 90K expression is not related to the We propose that in the context of the changing induced immune response but may reflect the TGF-β1-induced by TGF-β1 in the tumor microenvironment, the axis cellular microenvironment modifications that facilitate TGF-β1/USF/90K may take a part in the process of tumor cancer progression. In fact, experimental and clinical data progression. This possibility was further supported by our indicate that 90K is implicated in tumor invasion and analysis of the cBioPortal database showing a significant metastases, probably as a result of its ability to promote increased mortality in patients affected by ccRCC with cell-to-cell and cell-to-extracellular matrix (ECM) adhe- high levels of 90K expression in tumor samples. Similarly, sion9,37,38. Consistently, high expression levels of 90K, high levels of TGF-β1 or USF1 were associated with a both in serum or in tumor cell, have been observed in dismal prognosis. Moreover, a synergistic effect of 90K patients with different types of cancer and were asso- and TGF-β1 to predict higher mortality was observed. ciated with a higher incidence of metastasis and poor Importantly, in this cancer population, 90K levels directly – prognosis39 41. correlated with TGF-β1 and USFs expression.

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 12 of 13

Notably, the other gene activated by TGF-β1through References USF is PAI-1, a critical factor in the tumor invasion pro- 1. Morikawa, M., Derynck, R. & Miyazono, K. TGF-beta and the TGF-beta family: 51–57 β 58,59 context-dependent roles in cell and tissue physiology. Cold Spring Harb. Per- gram and implicated in the TGF- 1-induced EMT . spect. Biol. 8, https://doi.org/10.1101/cshperspect.a021873 (2016). The importance of TGF-β1 in the process of cancer 2. Batlle,E.&Massague,J.Transforming growth factor-beta signaling in immu- invasion and diffusion has led to the development of nity and cancer. Immunity 50,924–940 (2019). β 3. Wrzesinski, S. H., Wan, Y. Y. & Flavell, R. A. Transforming growth factor-beta and pharmacological inhibitors of the TGF- 1 signaling, and the immune response: implications for anticancer therapy. Clin. Cancer Res. 13, currently, some agents are in the early stages of clinical 5262–5270 (2007). trials60. PAI-1 has also been proposed as a target for 4. Viel, S. et al. TGF-beta inhibits the activation and functions of NK cells by 61 repressing the mTOR pathway. Sci. Signal 9,ra19(2016). cancer therapy . Similarly, we would propose 90K as 5. Chen, W. et al. Conversion of peripheral CD4+CD25- naive T cells to CD4 another possible anti-cancer target. +CD25+ regulatory T cells by TGF-beta induction of transcription factor In conclusion, the results of the present study on the Foxp3. J. Exp. Med. 198,1875–1886 (2003). β 6. Liu, Y. et al. A critical function for TGF-beta signaling in the development effect of TGF- 1 on 90K transcription and the role of of natural CD4+CD25+Foxp3+ regulatory T cells. Nat. Immunol. 9, USF1 and USF2 in the regulation of 90K gene expression 632–640 (2008). have uncovered a novel possible role for 90K in the 7. Gorelik, L. & Flavell, R. A. Transforming growth factor-beta in T-cell biology. Nat. β Rev. Immunol. 2,46–53 (2002). TGF- 1-driven tumor progression program. In this 8. Shull, M. M. et al. Targeted disruption of the mouse transforming growth regard,90Kmaybecomeasuitable candidate for anti- factor-beta 1 gene results in multifocal inflammatory disease. Nature 359, cancer therapy. 693–699 (1992). 9. Grassadonia, A. et al. 90K (Mac-2 BP) and galectins in tumor progression and metastasis. Glycoconj. J. 19,551–556 (2002). 10. Loimaranta, V., Hepojoki, J., Laaksoaho, O. & Pulliainen, A. T. Galectin-3-binding Acknowledgements protein: a multitask glycoprotein with innate immunity functions in viral and This work was supported by the Consorzio Interuniversitario Nazionale per la bacterial infections. J. Leukoc. Biol. 104,777–786 (2018). Bio-Oncologia (CINBO). Special thanks to Dina S. Singer (Experimental 11. Ullrich, A. et al. The secreted tumor-associated antigen 90K is a potent Immunology Branch, NCI, NIH, Bethesda, MD, USA) for her intellectual immune stimulator. J. Biol. Chem. 269,18401–18407 (1994). contribution. In memory of Leonard D. Kohn (NIDDK, NIH, Bethesda, MD, USA) 12. Natoli, C., Iacobelli, S. & Kohn, L. The immune stimulatory protein 90K increases fi who rst believed in this project. major histocompatibility complex class I expression in a human breast cancer cell line. Biochem. Biophys. Res. Commun. 225, 617–620 (1996). Author details 13. Kawashima, A. et al. Innate immune activation and thyroid autoimmunity. J. 1Department of Medical, Oral and Biotechnological Sciences and Center for Clin. Endocrinol. Metab. 96,3661–3671 (2011). Advanced Studies and Technology (CAST), G. D’Annunzio University, Chieti, 14. Grassadonia, A. et al. The 90K protein increases major histocompatibility Italy. 2Cancer Research UK Cambridge Institute, University of Cambridge, complex class I expression and is regulated by hormones, gamma-interferon, Cambridge, UK. 3Department of Medicine and Ageing Sciences and Center for and double-strand polynucleotides. Endocrinology 145, 4728–4736 (2004). Advanced Studies and Technology (CAST), G. D’Annunzio University, Chieti, 15. Napolitano, G. et al. Transforming growth factor-beta1 down-regulation of Italy. 4Division of Medical Oncology 2, IRCCS Regina Elena National Cancer major histocompatibility complex class I in thyrocytes: coordinate regulation Institute, Rome, Italy. 5Department of Pharmacy and Center for Advanced of two separate elements by thyroid-specificaswellasubiquitoustranscrip- Studies and Technology (CAST), G. D’Annunzio University, Chieti, Italy tion factors. Mol. Endocrinol. 14,486–505 (2000). 16. Pisarev, M. A., Thomasz, L. & Juvenal, G. J. Role of transforming growth factor – Author contributions beta in the regulation of thyroid function and growth. Thyroid 19,881892 A.G., V.G., and N.T. conceived the project. A.G., V.G., P.V., A.C., C.N., and N.T. (2009). designed the experiments and analyzed the data. S.P., A.V., C.G., M.M., P.L., 17. Carneiro, C., Alvarez, C. V., Zalvide, J.,Vidal,A.&Dominguez,F.TGF-beta1 M.D'E., P.D.M., and D.B. performed the experiments. A.G. and N.T. wrote the actions on FRTL-5 cells provide a model for the physiological regulation of – manuscript. M.D.T. and L.D.L. revised the manuscript. thyroid growth. Oncogene 16,1455 1465 (1998). 18. Ambesi-Impiombato, F. S., Parks, L. A. & Coon, H. G. Culture of hormone- dependent functional epithelial cells from rat thyroids. Proc. Natl Acad. Sci. USA Funding 77,3455–3459 (1980). This research received no external funding. 19. Veronese, A. et al. Mutated beta-catenin evades a microRNA-dependent regulatory loop. Proc. Natl Acad. Sci. USA 108,4840–4845 (2011). fl Con ict of interest 20. Grassadonia, A. et al. Upstream stimulatory factor regulates constitutive The authors declare no competing interests. expression and hormonal suppression of the 90K (Mac-2BP) protein. Endo- crinology 148,3507–3517 (2007). Ethical statement 21. Brakebusch, C. et al. Isolation and functional characterization of the human This research does not involve human subjects nor use animals. The origin of 90K promoter. Genomics 57, 268–278 (1999). all the cell lines is indicated in Material and Methods. 22. Giuliani, C. et al. Hormonal modulation of major histocompatibility complex class I gene expression involves an enhancer A-binding complex consisting of Fra-2 and the p50 subunit of NF-kappa B. J. Biol. Chem. 270, 11453–11462 Publisher’s note (1995). Springer Nature remains neutral with regard to jurisdictional claims in 23. Muller, S. A. et al. Domain organization of Mac-2 binding protein and its published maps and institutional affiliations. oligomerization to linear and ring-like structures. J. Mol. Biol. 291,801–813 (1999). Supplementary information The online version contains supplementary 24. Howcroft, T. K. et al. Upstream stimulatory factor regulates major histo- material available at https://doi.org/10.1038/s41420-021-00469-1. compatibility complex class I gene expression: the U2DeltaE4 splice variant abrogates E-box activity. Mol. Cell Biol. 19,4788–4797 (1999). 25. Riccio, A. et al. Transforming growth factor beta 1-responsive element: closely Received: 11 October 2020 Revised: 21 February 2021 Accepted: 18 March associated binding sites for USF and CCAAT-binding transcription factor- 2021 nuclearfactorIinthetype1plasminogen activator inhibitor gene. Mol. Cell Biol. 12,1846–1855 (1992).

Ofﬁcial journal of the Cell Death Differentiation Association Grassadonia et al. Cell Death Discovery (2021) 7:86 Page 13 of 13

26. Kutz, S. M. et al. TGF-beta 1-induced PAI-1 expression is E box/USF-dependent 44. Shariat, S. F. et al. Tissue expression of transforming growth factor-beta1 and and requires EGFR signaling. Exp. Cell Res. 312,1093–1105 (2006). its receptors: correlation with pathologic features and biochemical progres- 27. Freytag, J. et al. PAI-1 mediates the TGF-beta1+EGF-induced ‘scatter’ sion in patients undergoing radical prostatectomy. Urology 63,1191–1197 response in transformed human keratinocytes. J. Invest. Dermatol 130, (2004). 2179–2190 (2010). 45. Cantelli, G., Crosas-Molist, E., Georgouli, M. & Sanz-Moreno, V. TGFBeta-induced 28. Qi, L. et al. The basic helix-loop-helix/leucine zipper transcription factor USF2 transcription in cancer. Semin. Cancer Biol. 42,60–69 (2017). integrates serum-induced PAI-1 expression and keratinocyte growth. J. Cell 46. Ahmadi, A., Najafi,M.,Farhood,B.&Mortezaee,K. Transforming growth factor- Biochem. 115,1840–1847 (2014). beta signaling: Tumorigenesis and targeting for cancer therapy. J. Cell Physiol. 29. Shi,Y.&Massague,J.MechanismsofTGF-betasignalingfromcellmembrane 234, 12173–12187 (2019). to the nucleus. Cell 113, 685–700 (2003). 47. Bierie,B.&Moses,H.L.Tumourmicroenvironment: TGFbeta: the molecular 30. Macias, M. J., Martin-Malpartida, P. & Massague, J. Structural determinants of Jekyll and Hyde of cancer. Nat. Rev. Cancer 6,506–520 (2006). Smad function in TGF-beta signaling. Trends Biochem. Sci. 40,296–308 (2015). 48. Massague, J. TGFbeta in cancer. Cell 134,215–230 (2008). 31. Zawel, L. et al. Human Smad3 and Smad4 are sequence-specific transcription 49. Xu, J., Lamouille, S. & Derynck, R. TGF-beta-induced epithelial to mesenchymal activators. Mol. Cell 1, 611–617 (1998). transition. Cell Res. 19,156–172 (2009). 32. Itoh, Y. et al. A comparative analysis of Smad-responsive motifs identifies 50. McCormack, N. & O’Dea, S. Regulation of epithelial to mesenchymal transition multiple regulatory inputs for TGF-beta transcriptional activation. J. Biol. Chem. by bone morphogenetic proteins. Cell Signal 25,2856–2862 (2013). 294, 15466–15479 (2019). 51. Hirahata, M. et al. PAI-1, a target gene of miR-143, regulates invasion and 33. Inman, G. J. & Hill, C. S. Stoichiometry of active smad-transcription factor metastasis by upregulating MMP-13 expression of human osteosarcoma. complexes on DNA. J. Biol. Chem. 277, 51008–51016 (2002). Cancer Med. 5,892–902 (2016). 34. Zhou, S., Zawel, L., Lengauer, C., Kinzler, K. W. & Vogelstein, B. Characterization 52. Gomes-Giacoia, E., Miyake, M., Goodison, S. & Rosser, C. J. Targeting of human FAST-1, a TGF beta and activin signal transducer. Mol. Cell 2, plasminogen activator inhibitor-1 inhibits angiogenesis and tumor 121–127 (1998). growth in a human cancer xenograft model. Mol. Cancer Ther. 12, 35. Derynck, R. & Zhang, Y. E. Smad-dependent and Smad-independent pathways 2697–2708 (2013). in TGF-beta family signalling. Nature 425, 577–584 (2003). 53. Palmirotta, R. et al. Prognostic value of pre-surgical plasma PAI-1 (plasminogen 36. Massague, J. TGFbeta signalling in context. Nat. Rev. Mol. Cell Biol. 13,616–630 activator inhibitor-1) levels in breast cancer. Thromb. Res. 124,403–408 (2009). (2012). 54. Halamkova, J. et al. Clinical relevance of uPA, uPAR, PAI 1 and PAI 2 tissue 37. Sasaki, T., Brakebusch, C., Engel, J. & Timpl, R. Mac-2 binding protein is a cell- expression and plasma PAI 1 level in colorectal carcinoma patients. Hepato- adhesive protein of the extracellular matrix which self-assembles into ring-like gastroenterology 58,1918–1925 (2011). structures and binds beta1 integrins, collagens and fibronectin. EMBO J. 17, 55.Ferroni,P.etal.Plasmaplasminogenactivator inhibitor-1 (PAI-1) levels in 1606–1613 (1998). breast cancer - relationship with clinical outcome. Anticancer Res. 34, 38. Ulmer, T. A. et al. The tumor-associated antigen 90K/Mac-2-binding protein 1153–1161 (2014). secreted by human colon carcinoma cells enhances extracellular levels of 56. Chen, H. et al. Silencing of plasminogen activator inhibitor-1 suppresses promatrilysin and is a novel substrate of matrix metalloproteinases-2, −7 colorectal cancer progression and liver metastasis. Surgery 158, (matrilysin) and −9: Implications of proteolytic clea. Biochim. Biophys. Acta 1704–1713 (2015). 1800,336–343 (2010). 57. Fredstorp-Lidebring, M. et al. Urokinase plasminogen activator and its inhibitor, 39. Iacobelli, S. et al. Prognostic value of a novel circulating serum 90K antigen in PAI-1, in association with progression-free survival in early stage endometrial breast cancer. Br.J.Cancer69,172–176 (1994). cancer. Eur. J. Cancer 37,2339–2348 (2001). 40. Marchetti, A. et al. Expression of 90K (Mac-2 BP) correlates with distant 58. Omori, K. et al. Inhibition of plasminogen activator inhibitor-1 attenuates metastasis and predicts survival in stage I non-small cell lung cancer patients. transforming growth factor-beta-dependent epithelial mesenchymal transi- Cancer Res. 62, 2535–2539 (2002). tion and differentiation of fibroblasts to myofibroblasts. PLoS ONE 11, 41. Tinari, N. et al. High expression of 90K (Mac-2 BP) is associated with poor e0148969 (2016). survival in node-negative breast cancer patients not receiving adjuvant sys- 59. Thiery, J. P. Epithelial-mesenchymal transitions in tumour progression. Nat. Rev. temic therapies. Int J. Cancer 124,333–338 (2009). Cancer 2,442–454 (2002). 42. Elliott, R. L. & Blobe, G. C. Role of transforminggrowthfactorBetainhuman 60. Chen, Y. et al. Transforming growth factor beta signaling pathway: a promising cancer. J. Clin. Oncol. 23, 2078–2093 (2005). therapeutic target for cancer. J. Cell Physiol. 235, 1903–1914 (2020). 43. Narai,S.etal.Significance of transforming growth factor beta1 as a new 61. Li, S. et al. Plasminogen activator inhibitor-1 in cancer research. Biomed. tumor marker for colorectal cancer. Int J. Cancer 97,508–511 (2002). Pharmacother. 105,83–94 (2018).

Ofﬁcial journal of the Cell Death Differentiation Association